convo-sdk - NFHN Reader

🧠 Convo SDK — Persistent Memory for LangGraph Agents

LLMs forget everything.
Convo remembers — so your agents don’t have to.

Convo is a drop-in LangGraph checkpointer that gives your AI agents long-term memory, workflow traceability, and easy debugging.

It works out of the box with LangGraph. One import = state persistence for multi-agent setups.

import { Convo } from "convo-sdk";
const convo = new Convo({ apiKey: "your-convo-api-key" });

const thread_id = await convo.newThread(); // Or reuse an existing one

const checkpointer = convo.checkpointer({ thread_id });

const graph = workflow.compile({ checkpointer });
await graph.invoke(state, {
  configurable: { thread_id }
});

await graph.invoke(state, {
  configurable: {
    thread_id,
    checkpoint_id: "your-checkpoint-id",
  },
});

Method	Description
`await convo.init({ apiKey })`	Initialize SDK
`await convo.newThread()`	Create a conversation thread
`convo.checkpointer({ thread_id? })`	Get LangGraph-compatible checkpointer
`graph.invoke(state, { configurable })`	Run with checkpointing + config
`convo.getThreadState(thread_id)`	Fetch full state of a thread
`convo.listThreads()`	View all your conversation threads

You can either:

Set at checkpointer creation
Persistent across runs:

const checkpointer = convo.checkpointer({ thread_id });

Or pass dynamically during graph invocation
Per-run control:

await graph.invoke(state, {
  configurable: { thread_id }
});

Convo is ideal for:

With just one import, Convo:

Build smarter agents. Persist their thoughts.
Convo makes stateful AI as simple as stateless.